基于高光谱结合机器学习方法对谷子株高的遥感监测OA

To develope high-accuracy models for foxtail millet plant height remote monitoring,this study used foxtail millet variety'Jingu 21 hao'as the test material.Canopy hyperspectral data were collected with an ASD FieldSpec spectrometer,and corresponding plant height measurements were taken simultaneously.Five spectral preprocessing methods-including first derivative,second derivative,standard normal variate,multiplicative scatter correction,and Savitzky-Golay smoothing-were applied to the original reflectance(R)data.This study then combined vegetation indices,characteristic wavelengths,and full-band data with four machine learning algorithms to explore their correlations with plant height and construct a monitoring model.The results showed that the optimal vegetation index(NPQI)correlated with plant height at a coefficient of-0.718.Among the four machine learning models developed,the 2ST-PLS model(R2=0.850,RMSE=6.655 cm,RPD=2.187),based on full-spectrum data across all growth stages,and the PLS model(R2=0.840,RMSE=6.102 cm,RPD=2.385)exhibited strong predictive capability.This study demonstrates the feasibility of using spectral techniques for monitoring foxtail millet plant height and provides a methodological basis for its precise,rapid,and non-destructive estimation.